Facebook researchers detail progress in contextual chatbots for shopping

Brief:

Facebook researchers last week described how data about human conversations can lead to improvements in shopping chatbots, according to a technical report shared by Venture Beat.
In the report about Situated Interactive MultiModal Conversations (SIMMC), researchers demonstrated how information gleaned from 169,000 human utterances in 13,000 discussions were used to automate conversations about shopping for furniture and fashion.
User satisfaction for the conversations about furniture shopping averaged about 4.7 on a scale of 1 to 5, indicating that people were generally happy with the chatbot's results. Facebook also saw a similar trend for the data gleaned from conversations about fashion, per the report.

Insight:

Facebook's research into shopping chatbots suggests the technology can be greatly improved with more data from conversations that live humans have with each other about products such as furniture and fashion. It's too early to tell how the social media giant will apply the advancements in natural-language processing (NLP) to real-world scenarios, but the research signals that the technology can provide a positive customer experience (CX). For brands and retailers seeking to provide better customer service while reducing the demands on call centers, the improved shopping chatbot technology may be significant.

For mobile marketers, chatbots' sophistication can help to drive online sales while saving on labor costs for call centers. Consumer spending through chatbots is forecast to surge to $142 billion by 2024 from $2.8 billion in 2019, Juniper Research estimated. Discrete chatbots — technology embedded directly into a retailer's mobile app rather than accessed through a browser or messenger — are forecast to generate 80% of global spending through chatbots by 2024. The research suggests retailers will rapidly adopt chatbot tech in the next few years.

The research included ways to generate chatbot responses that make sense and consider the context. A shopper could request, "Show me black couches less than $500," and see responses such as "Here are some" or "Sorry, we do not have any black couches cheaper than $500," depending on their inventory availability, as an example cited in the report. That suggests that chatbots can be upgraded to improve customer satisfaction, a welcome development for a technology that has been derided for not understanding how people talk in real life. Only 16% of consumers said they had a good experience with a chatbot, according to a 2019 survey by Drift and SurveyMonkey, compared with the 39% who said they had a good experience with an online chat platform.

It's unclear how Facebook will apply its research into shopping chatbots among the merchants with digital storefronts on its apps. CEO Mark Zuckerberg four years ago touted chatbots as a promising technology to automate customer service and e-commerce transactions. His announcement led to speculation that the company wanted to emulate Tencent's WeChat, the Chinese app that handles dozens of functions with "mini programs" for everything from online shopping to bill payments.

However, the company this year revamped its Messenger chat app to focus on Stories while de-emphasizing chatbots as part of a redesign. Facebook's ongoing research into shopping chatbots suggests that it still sees potential for the technology to support e-commerce, a key area of focus for the company in diversifying its revenue beyond advertising.